Current estimates are that there’s a roughly 1 in 1000 difference between two individuals purely due to random mutations in the germline (or, put differently, the DNA replication error rate in human cells is about 1 in 1000). This would imply 99.9% expected similarity between two identical twins. However, this is in the germline. Somatic cells accumulate mutations over their life time (since they, too, divide and therefore need to replicate their DNA, and are furthermore exposed to environmental stress). In addition, the tests work off saliva samples which are mixed with all kinds of stuff in the mouth, and degrade slightly during transport. Lastly, there’s an inherent replication variability in the genotyping process (though this is very small).
This back of the envelope estimate suggests that 99.6% similarity for somatic samples of identical twins is not unexpected.